Applied Data Science UNBOXED
In each episode of Applied Data Science UNBOXED, we unpack a topic and show how data science creates impact across a wide range of fields. Host Fabio Sandmeier explores concrete applications of data science with experts from academia, industry, and society.
This podcast is part of the MSc in Data Science and AI at the Lucerne University of Applied Sciences and Arts (HSLU).
Can’t get enough of data science?
Applied Data Science UNBOXED
From thesis to publication – How smartwatches could transform diabetes care
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Episode 9 | How can data science improve the lives of people with diabetes? In her master’s thesis, Yasmine Mohamed explored whether everyday smartwatches can reliably detect hypoglycemia – and the results were surprisingly promising.
Supported by her supervisor, José Mancera, she pushed the project far beyond the boundaries of a typical thesis: from ethics approval and data collection to publication in a scientific journal.
00:11 – Motivation: Why Yasmine chose diabetes and her personal drive
05:26 – The Idea: Smartwatches as an alternative to medical-grade devices
08:15 – Challenges: Ethics approval, funding, and the long wait for data
12:09 – Breakthrough: First data, first insights, and the “Eureka moment”
15:21 – The Supervisor’s View: How José is guiding, not leading
19:18 – Results & Outlook: Smartwatches detecting hypoglycemia – and Yasmine’s next steps in research
- Read the publication.
- More about Yasmine Mohamed and her Master's Thesis
- More about José Mancera
Applied Data Science UNBOXED – Knowledge. Trends. Future. We open the door to the world of data science and show how data is shaping our future.
Whether you're a data science enthusiast or a career changer, you'll find real insights here. Renowned experts from science, industry, and society talk about current trends, groundbreaking innovations, and the most exciting challenges in the field.
Curious? Then tune in and immerse yourself in the world of data. Learn more:
🔹 Master's program in Applied Information and Data Science – all information online
🔹 Free online info events – directly from the program directors
🔹 Our blog – current articles on data science
🔹 LinkedIn – network and never miss an update
About the podcast
In each episode of Applied Data Science UNBOXED, we unpack a topic and show how data science creates impact across a wide range of fields. Host Fabio Sandmeier explores concrete applications of data science with experts from academia, industry, and society.
This podcast is part of the MSc in Applied Data Science and AI at the Lucerne University of Applied Sciences and Arts (HSLU).
In this episode, we will talk about diabetes and a student who set out to make life easier for diabetes patients.
SPEAKER_01My main drive is that I don't want to do something just to graduate. I want to do something impactful. It sounds a bit cliche, but I want to do something that would actually have real life implementation and value.
SPEAKER_02Welcome to Applied Data Science Unboxed. My name is Fabio Sandmeyer, and in this podcast, we look behind the term data science. What does it look like in practice? Today we're talking about Yasmin Mohamed's master thesis. She was looking for new ways to help diabetes patients and was supported by her lecturer José Mancera.
SPEAKER_00So I'm here as a facilitator, I'm just yes, I'm a supervisor. But the idea here, because you're in your end of your studies, is you have to lead this, you are the tech lead here, you are the project lighter or project manager of this, right? And normally I say these initial words, and then she took it really seriously.
SPEAKER_02I wanted to find out what makes an effective master's thesis. How do you set up a project that goes beyond the usual boundaries? And what mindset do you need for it? Diabetes is one of the most common chronic conditions worldwide, and there's still no cure. Managing it means finding the right balance of diet, exercise, and medication. Before this podcast episode, I was not aware of the dimensions of diabetes. That's why I want to give you first some numbers. Worldwide, the number of people with diabetes rose from around 200 million to 830 million over the last 35 years. In Switzerland, it is estimated that around 525,000 people suffer from diabetes. This corresponds to around 7% of the population. For people with type 1 diabetes, low blood sugar episodes known as hypoglycemia are one of the biggest challenges. They don't just increase the risk of accidents or even death, they also create a constant psychological burden because hypoglycemia can strike unpredictably. Many people live with a fear for the next episode, which affects their quality of life. That's why monitoring blood sugar is so important. Today, this is mostly done with fingerpricks or continuous glucose monitoring systems. These devices have made life easier by providing real-time feedback. But in everyday life, they can still be inconvenient, and that's why we need additional ways to better support diabetes management. Yasmin Mohamed is a graduate of the Master of Applied Information and Data Science. She tackled this topic in her master's thesis and she went beyond the boundaries of a standard master's thesis. I meet Yasmin Mohamed at the computer science campus of Ulcerne University of Applied Science and Arts in Rotkreuz.
SPEAKER_01It all started in a module. So we were supposed to come up with an idea for a startup company to do this project work. It's actually one of the best modules I've attended, and this is where I got to know Jose. So I came up, given my background in healthcare, I suggested that this startup be a healthcare company that is trying to come up with smart solutions using technology, things like wearables, et cetera, to help people manage their illnesses, specifically chronic diseases. I think maybe diabetes was some sort of subconscious decision, I guess. I mean I've had personal experience. I'm not a not a person with diabetes, but I've known many people in my family. I've also had a childhood friend who had type 1 diabetes, and I got to experience firsthand how challenging this disease actually is on a day-to-day basis. And I remember once I was on the phone with this friend, and they just woke up and they had an episode of low blood sugar. It's called hypoglycemia. And I remember they were so confused, the like slurred speech, not really oriented. I think it really stuck in my head how challenging it can be. The thing with diabetes is that it's not like you're just replaced, it's not just hormone replacement. That's it sounds simple, but it's not simple because insulin, the hormone that is usually missing in the body of uh someone who has type 1 diabetes, it's uh released on demand. It's not something that just has a certain rhythm in the body. So a person with diabetes would have to balance their diet, their exercise, and their insulin shots. It's not easy. So I think maybe that was the drive behind choosing diabetes for this project. There is actually wearable technology that is starting, I think maybe you've seen someone wearing a continuous glucose monitor on their arm or something. But going back to this project that was a requirement for the module, we were supposed to create a data pipeline where we would collect data and store it and then create a dashboard to visualize it. So we were collecting data on publications, on clinical trials, et cetera. And I started to come across these publications that were using wearable technology, however, medical grade, so things like ECG equipment, et cetera, to measure the physiological changes that happen in the body of someone who has diabetes when their glucose levels or their blood sugar levels are not in their ideal state. So they're too high or too low. The focus is specifically on the lows because the consequences are immediate. So a person who has an episode of hypoglycemia or low blood sugar would immediately start to be confused, sweating, shaking, etc. And if it gets too low, then this is an emergency situation. It's dangerous. Yeah, so I was it really sparked my interest, these papers. But I noticed, like I told you, that they're all medical grade. And even the ones that are we would say are wearables that could be used when someone is doing their, you know, going about their life naturally. These were chest straps that you would not wear a chest strap every day, but you would wear a smartwatch every day. You would go swimming in it, have a shower, sleep on you all the time. So I was curious about whether has someone have explored this domain, but using a wearable that is commercially available to people, something that people would actually use and there wasn't any. So that's when the idea came to mind. And I had to approach supervisors. Like I told you, I got to know Jose during this module. So I approached him as a supervisor, and he was just great, very supportive, got so excited about the idea, and that's I guess that's how it started.
SPEAKER_02What followed was a project that went beyond the boundaries of the usual master's thesis. Yasmin Mohammed approached hospitals and looked for project partnerships, but there were also a few hurdles to overcome. You remember the last episode with ethicist Orlando Butelacci. She also had to deal with his committee, the ethics committee, because when it comes to issues like this, you look closely at the ethical level.
SPEAKER_01The main challenge was the time frame. From the very start of the conception of this idea, the main challenge was time. I had to approach Jose, but honestly, he was so supportive, so quick to get on board. And then we needed funding because we had to buy the Apple Watches and so on. And then I had to find another co-supervisor. And I found Dr. Michael Havranek was also big support from the University of Luzern. And then he, because he has experience, he's the head of the competence center of health data science. So he has more experience and he started to point me into what are the next steps. So the first thing he told me was that you need to get the approval of the ethics committee because we're doing research with individuals, with people, humans. So we have to get their approval. And this process can take up to six months. And if this demotivates you, then I advise you to look for another idea. And then I was like, no, I'm not demotivated. I want to go on with it. And then he started to connect me with Dr. Stefan Fischley, the head of the endocrinology department in the Lucana Canton Schwitat. And then we started to work on the on the application to the ethics committee. And it did take six months.
SPEAKER_02It takes six months.
SPEAKER_01Yes. Although the ethical considerations are not very high because we're just going to give the participants an Apple Watch and a continuous glucose monitor. These are two wearables. They're not collecting any sensitive data. It's just physiological readings and motion and blood sugar. Nothing that will violate their privacy. Nonetheless, we need to get their approval. Then we had to think about the study design. This is also one of the of the main issues that can raise ethical issues. For example, the study design we wanted the participants to be blinded with their continuous glucose monitors, so not to see their blood sugar levels. For people, for example, who were used to using a CGM, this might have been a little bit problematic. You're taking away from them a tool that they are using. But this was also a minor issue. Although participants consented, there was no problem. The main challenge was the time frame. But also Professor Brandenberg was so supportive with the funding with extending the deadline for submitting the thesis. So once I got the green light that you can take however long time you want, I was not concerned anymore.
SPEAKER_02So did this then prolong your studies?
SPEAKER_01Absolutely, yes. And then after waiting these six months, we had to start with the data collection. We had no data. And that took a year. Whether it was a good idea, no. Whether it was taking so long, yes. But I was motivated. And once you know the data started coming in, I didn't wait until all participants joined. I started to work already on the data, and once I started receiving the data and working on it, this spiked my motivation.
SPEAKER_02For a long time I thought that data science meant analyzing data. But with Yasmin's project, you can really feel that data science is largely about obtaining usable data. It's digging for gold. And then the moment when the data is available, that's the first summit moment. The second summit moment is then the first insight. And in Yasmin's case, it was going to take a while until then. How was that feeling when you saw the data coming in?
SPEAKER_01Oh, it was fantastic. I mean, I had also a pipeline, like while the participant was wearing the watch, I had a way to actually receive it in a what you call it, uh using Amazon Web Services. They have something called an S3 bucket where you can store the data. So I was already seeing the data coming through, and that was such a great feeling. I've been waiting so long with getting the funding and the approvals and so on, and then this moment just is actually happening, you know? Yeah, a mixture of anxious and excited.
SPEAKER_02But yeah. And how long did it take you then from seeing the data till you got some first insights?
SPEAKER_01A couple of months, I would say. Because also the data itself, like the nature of the data itself was challenging. So I had to do many steps of transformation before I can actually use it. And then the very nature of the data it required custom methodology to actually tackle it and get results. I would say b between actually receiving the data and getting some insights, it took another three, four months.
SPEAKER_02And what was then the moment when you got the first insight, the first result? Was this in form of a specific diagram or how did this look like?
SPEAKER_01What we were trying to do is to see whether we are collecting this data from two devices. The continuous glucose monitor, which is measuring sugar levels every five minutes, and the Apple Watch that is continuously measuring physiological parameters like heart rate variability, etc. So what we're trying to do is to get this data, give part of it to an algorithm, a machine learning model, and then say, okay, can you try to learn from this data that I'm giving you a pattern so that when I show you new data, you would be able to recognize an episode of hypoglycemia from the physiological from the changes in the physiological parameters. So the insight that I'm trying to get is a good model performance. The model adequately has a good level of performance that it can actually detect these episodes of hypoglycemia. And of course, the first time this happened, it was a Eureka moment. I mean, it was so exciting. Yeah.
SPEAKER_02Where were you in that moment?
SPEAKER_01At home. Yeah, yeah. Yeah, it was I was so excited, and I started like contacting Jose and Michael, like, I'm getting results. Yeah.
SPEAKER_02Yeah, so exciting. But the data from the Apple Watches, they were able to tell quite accurately when someone's blood sugar was going too low. I also spoke to Jose Mancera, who was mean cold in this Eureka moment.
SPEAKER_00So hello, my name is Jose Mancera, and I'm a lecturer in the Master of Data Science. I originally become from the EKM, so Institute of Communications Marketing. So this is where I'm fully assigned. However, I have a lot of experience with machine learning, and I currently give two modules. One is about data visualization, and the other one is data lake and data warehouses with a little bit more data engineering. So my background in general is more technical, let's say, more like applied data science.
SPEAKER_02I meet Jose in a corner of the canteen at the Lucerne School of Business. It's lunchtime, it's busy, but Jose is not faced by this. With a smile on his lips and the gleam in his eye, he talks about data science and especially about the project with his former student Yasmin Mohammed.
SPEAKER_00So Jasmine, she was deeply interested in everything that is related to medicine. And for a certain direction, she was just knocking doors and she contacted a professor at the University of Lucerne. Then they he was offering like very theoretical topics. Theoretical topics in terms of medicine. Like she was very interested in this particular line. But then there was an opportunity that this professor was for the very first time wanted to collaborate with the Cantonal Hospital Lucerne, right? With the diabetes research group. That was actually a very interesting opportunity because they were very, very theoretical based in the University of Lucerne. But in the HaselU, we are very practice-based. So I think it was kind of luck, and also like she was knocking doors everywhere, and she was able to manage to get this project where she can use her data science skills related to a field that she's interested in, which is medicine, right? Or in this case, like diabetes detection, there was something that it was kind of like several factors there, but definitely it was more the drive from the student to find these projects, searching for something in medicine. I think that was she was taking a lot of initiative on this. Yeah.
SPEAKER_02And then throughout the course of the project, how was the collaboration?
SPEAKER_00So it was very interesting because when we have our first kickoff meeting, first I have one with her, and then I told her, like, you know, I mean, this is your project. So I'm here as a facilitator. I'm just I mean, yes, I'm your supervisor. But the idea here, because you are in your end of your studies, is you have to lead this, you are the tech lead here, you are the project lighter or project manager of this, right? And normally I say these initial words, and then she took it really seriously. So she really took the ownership of the project. So she was coordinating meetings and she was like doing like this planning for the working packages for not only for her thesis here at Hasseloup, but also like what are the working packages that they have to do in research at the University of Lucerne? Also, like when is the coordination of patients as well with doctors? That was actually very impressive. That's what she was like really taking the ownership of the project as a project manager.
SPEAKER_02Once more, back to Yasmin. She did not simply follow instructions or took the easy route. She scrutinized existing methods. I ask myself, where does this mindset come from?
SPEAKER_01Hmm. I come from Egypt. And situations are not always ideal. You always have to think of a workaround to make things work. This is somehow where this is coming from. And in my upbringing, I was not brought up to just accept the status quo as is you have to question, even if this is something that has been going on for ages. You get to question it and do whatever you think is the right thing.
SPEAKER_02You're probably wondering what came out of the project. Their study shows that everyday smartwatches can help detect low blood sugar in people with type 1 diabetes without the need for needles or invasive devices. Using machine learning on smartwatch data, they were able to predict hypoglycemia with fairly good accuracy. A few weeks ago, their study was published in a prestigious scientific journal. You can find the link in the show notes. Now it remains to be seen who in science will pick up the thread and continue the research or develop it further. For Yasmin Mohammed, the master's thesis was a stepping stone to an exciting position in the pharmaceutical industry. She works now as a research fellow in a health data science project where she helps to better recognize tropical diseases such as malaria or dengue fever. As you can see, Yasmin remains true to herself. You can find out more about Yasmin Mohammed, her master's thesis, and of course the master IDS in the show notes below. My name is Fabia Sandmeyer. I hope you like this episode. This was a podcast from the Lucerne University of Applied Sciences and Arts. Thank you for listening and see you next time.